Vincent Purcell

Data Mining 2 - Final Project

Set up MongoDB Connections

import pymongo

# MongoDB Connections
client = pymongo.MongoClient('mongodb://localhost:27017')
db = client['MLB_DB']

Stat Averages Function

Function to take array of statistics and get average of each feature and output list of averages

def getStatAvg(stats_list, projections_str, player_projected_stats_doc):
    stat_averages = [0] * len(stats_list)
    count = 0
    for player in player_projected_stats_doc[projections_str].values():
        stat_index = 0
        for stat in stats_list:
            try:
                stat_val = float(player[stat])
            except:
                stat_val = 0
            stat_averages[stat_index] = stat_averages[stat_index] + stat_val
            stat_index = stat_index + 1
        count = count + 1
    stat_averages = [i/count for i in stat_averages]
    return stat_averages

Get Data Function

Function to grab data from specified DBs and output an array

import datahelp

def getData(summaryDB, PlayerDataDB, PlayerProjectionsDB):
    columns = datahelp.columns
    data = []
    for document in summaryDB.find():
        game_id = document['game_id_num']
        for player_data in PlayerDataDB.find({'game_id_num': str(game_id)}):
            home_umpire = [player_data['umpires']['home']['id']]
        for player_projections in PlayerProjectionsDB.find({'game_id_num': str(game_id)}):
            # Get Stat averages
            home_batting_avg = getStatAvg(datahelp.batting_stats, 'home_batting_projections', player_projections)
            away_batting_avg = getStatAvg(datahelp.batting_stats, 'away_batting_projections', player_projections)
            home_pitching = getStatAvg(datahelp.pitching_stats, 'home_pitcher_projections', player_projections)
            away_pitching = getStatAvg(datahelp.pitching_stats, 'away_pitcher_projections', player_projections)
        home_win = [1 if document['home_score']>document['away_score'] else 0]
        data_row = home_batting_avg + away_batting_avg + home_pitching + away_pitching + home_umpire + home_win
        data.append(data_row)
    return data

Read data from DB's

gameSummaryDB = db['GameSummary']
gamePlayerDataDB = db['GamePlayerData']
playerProjectionsDB = db['PlayerProjectedStats']

data = getData(gameSummaryDB, gamePlayerDataDB, playerProjectionsDB)

Create Pandas Dataframe

import numpy as np
import pandas as pd
import datahelp

columns = datahelp.columns

data_df = pd.DataFrame(np.array(data), columns=columns)
print("Data count (rows, columns) : " + str(data_df.shape))
data_df.head()

Data count (rows, columns) : (1202, 34)

	home hr avg	home rbi avg	home tb avg	home avg avg	home ops avg	home so avg	home tpa avg	home obp avg	away hr avg	away rbi avg	...	away pitching hr	away pitching wpct	away pitching era	away pitching outs	away pitching whip	away pitching h	away pitching ip	away pitching pa	umpire_id	home win
0	15.222222	80.666667	283.111111	0.259111	0.715444	160.888889	800.333333	0.321556	29.000000	117.666667	...	14.0	0.667	3.67	427.0	1.41	158.0	142.1	617.0	484499.0	0.0
1	16.777778	77.444444	273.888889	0.250778	0.701222	158.555556	768.888889	0.316111	30.222222	102.888889	...	9.0	0.167	5.46	173.0	1.63	70.0	57.2	265.0	482666.0	0.0
2	16.333333	63.222222	214.555556	0.230000	0.673111	130.444444	586.777778	0.295556	25.666667	83.777778	...	35.0	0.318	6.10	575.0	1.57	226.0	191.2	869.0	427315.0	0.0
3	18.555556	78.222222	265.444444	0.254111	0.722778	158.555556	720.000000	0.316111	23.555556	89.222222	...	46.0	0.478	3.89	1166.0	1.34	371.0	388.2	1654.0	428442.0	1.0
4	17.555556	81.444444	280.888889	0.262444	0.728667	150.222222	752.888889	0.319111	27.111111	86.222222	...	30.0	0.429	4.86	766.0	1.54	264.0	255.1	1143.0	427044.0	1.0

5 rows × 34 columns

Get Training and Testing Data

Perform 85/15 test split. Print data snippet and class count plot

from sklearn.model_selection import train_test_split
import seaborn as sns
import matplotlib.pyplot as plt

X = data_df.drop(['home win'], axis = 1)
y = data_df['home win']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.15, random_state = 15)
# visualize number of digits classes
plt.figure(figsize=(5,5))
g = sns.countplot(y_train, palette="icefire")
plt.title("Number of Wins and Losses")
y_train.value_counts()

X_train.head()

Pass the following variable as a keyword arg: x. From version 0.12, the only valid positional argument will be `data`, and passing other arguments without an explicit keyword will result in an error or misinterpretation.

	home hr avg	home rbi avg	home tb avg	home avg avg	home ops avg	home so avg	home tpa avg	home obp avg	away hr avg	away rbi avg	...	home pitching pa	away pitching hr	away pitching wpct	away pitching era	away pitching outs	away pitching whip	away pitching h	away pitching ip	away pitching pa	umpire_id
976	22.222222	83.444444	273.555556	0.241889	0.693444	132.555556	755.111111	0.304667	42.333333	126.888889	...	816.0	15.0	0.625	4.12	275.0	1.13	85.0	91.2	377.0	521251.0
136	37.666667	113.777778	357.444444	0.264889	0.813000	182.777778	828.888889	0.329889	34.222222	121.555556	...	1559.0	4.0	0.600	5.30	107.0	1.43	34.0	35.2	162.0	503493.0
1092	11.000000	35.666667	117.444444	0.220444	0.663444	85.666667	341.111111	0.292222	11.111111	41.666667	...	371.0	9.0	0.600	3.48	155.0	1.20	43.0	51.2	215.0	427139.0
981	28.000000	100.222222	313.000000	0.253667	0.770778	190.888889	800.000000	0.327778	23.333333	82.000000	...	1723.0	57.0	0.344	5.72	699.0	1.40	252.0	233.0	1024.0	547380.0
620	33.777778	129.000000	393.555556	0.262778	0.770889	171.888889	1001.777778	0.336667	22.666667	84.555556	...	1655.0	28.0	0.385	4.68	629.0	1.37	214.0	209.2	912.0	427128.0

5 rows × 33 columns

Correlation Heatmap

sns.heatmap(X_train.corr()).set_title("Data Heatmap");

Data Pairplot

sns.pairplot(data_df, hue='home win', vars=['home pitching outs', 'away pitching outs','home rbi avg', 'away rbi avg']);

MLP Classifier

Output MLP accuracy and confusion matrix

from sklearn.neural_network import MLPClassifier
from sklearn.metrics import classification_report, confusion_matrix

mlp = MLPClassifier(hidden_layer_sizes=(50,75,98), max_iter=2000)
mlp.fit(X_train,y_train)
y_predict = mlp.predict(X_test)

cm = np.array(confusion_matrix(y_test, y_predict, labels=[1,0]))
confusion = pd.DataFrame(cm, index=['home_win', 'home_loss'],
                         columns=['predicted_win','predicted_loss'])
sns.heatmap(confusion, annot=True)
plt.title("MLP Confusion Matrix");

SVM Classifier

Output SVM accuracy and confusion matrix

from sklearn.svm import SVC
svc_model = SVC()
svc_model.fit(X_train, y_train)
y_predict = svc_model.predict(X_test)

cm = np.array(confusion_matrix(y_test, y_predict, labels=[1,0]))
confusion = pd.DataFrame(cm, index=['home_win', 'home_loss'],
                         columns=['predicted_win','predicted_loss'])
sns.heatmap(confusion, annot=True)
plt.title("SVM Confusion Matrix");

SVM Regressor Sensitivity Analysis

import shap
from sklearn.svm import SVR
svr_poly = SVR(kernel='poly', C=1e3, degree=2)
svr_poly.fit(X_train, y_train)
y_predict = svr_poly.predict(X_test)
explainer = shap.explainers.Permutation(svr_poly.predict, X_test)
shap_values_svc = explainer(X_test[:100], silent=True)
shap.plots.bar(shap_values_svc)

MLP Regressor Sensitivity Analysis

from sklearn.neural_network import MLPRegressor
mlp_regr = MLPRegressor(random_state=1, max_iter=2000).fit(X_train, y_train)
y_predict = mlp_regr.predict(X_test)
explainer = shap.explainers.Permutation(mlp_regr.predict, X_test)
shap_values_svc = explainer(X_test[:100], silent=True)
shap.plots.bar(shap_values_svc)

Decision Tree Regressor Sensitivity Analysis

from sklearn.tree import DecisionTreeRegressor
dt_regr = DecisionTreeRegressor(max_depth=7)
dt_regr.fit(X_test,y_test)
explainer = shap.explainers.Permutation(dt_regr.predict, X_test)
shap_values_svc = explainer(X_test[:100], silent=True)
shap.plots.bar(shap_values_svc)

Decision Tree Classifier (Max depth specified)

from sklearn import tree
from dtreeviz.trees import *
import dtreeviz as dtviz


max_depth = 3
clf_dt = tree.DecisionTreeClassifier(max_depth=max_depth)
clf_dt = clf_dt.fit(X_train, y_train)
#plt.figure(figsize=(35,12))
#tree.plot_tree(clf_dt,feature_names=columns,class_names=['loss','win'],rounded=True,fontsize=8);
#plt.show()
predictions_dt = clf_dt.predict(X_test)
accuracy = 1-(((y_test != predictions_dt).sum())/len(X_test))
print("Decision Tree Classification (max depth =" + str(max_depth) + ")")
print("Number of mislabeled points out of a total %d points: %d" % (len(X_test), (y_test != predictions_dt).sum()))
print("Accuracy: %f" % (accuracy))



viz = dtviz.trees.dtreeviz(clf_dt,
                   X_train,
                   y_train,
                   target_name='home_win',
                   feature_names=columns[:-1],
                   orientation="TD",
                   class_names=['loss','win'],
                   fancy=True,
                   X=None,
                   label_fontsize=12,
                   ticks_fontsize=8,
                   fontname="Arial")

viz

Decision Tree Classification (max depth =3)
Number of mislabeled points out of a total 181 points: 56
Accuracy: 0.690608

Decision Tree Classifier (No Max Depth)

clf_dt = tree.DecisionTreeClassifier()
clf_dt = clf_dt.fit(X_train, y_train)
plt.figure(figsize=(100,15))
tree.plot_tree(clf_dt,feature_names=columns,class_names=['loss','win'],rounded=True,fontsize=7);
#plt.savefig('dtree_high_dpi',dpi=600)
predictions_dt = clf_dt.predict(X_test)
accuracy = 1-(((y_test != predictions_dt).sum())/len(X_test))
print("Decision Tree Classification")
print("Number of mislabeled points out of a total %d points: %d" % (len(X_test), (y_test != predictions_dt).sum()))
print("Accuracy: %f" % (accuracy))

Decision Tree Classification
Number of mislabeled points out of a total 181 points: 75
Accuracy: 0.585635

Random Forest Classifier

from sklearn.ensemble import RandomForestClassifier

max_depth = 7
criterion = "entropy"
clf_rf = RandomForestClassifier(max_depth=max_depth, random_state=0, criterion="entropy", n_estimators=150)
clf_rf.fit(X_train, y_train)
predictions_rf = clf_rf.predict(X_test)

accuracy = 1-(((y_test != predictions_rf).sum())/len(X_test))
print("Random Forest Classification")
print("max depth = " + str(max_depth) + ", criterion = " + criterion)
print("Number of mislabeled points out of a total %d points: %d" % (len(X_test), (y_test != predictions_rf).sum()))
print("Accuracy: %f" % (accuracy))

estimator = clf_rf.estimators_[5]

viz = dtviz.trees.dtreeviz(estimator,
                   X_train,
                   y_train,
                   target_name='home_win',
                   feature_names=columns[:-1],
                   orientation="LR",
                   class_names=['loss','win'],
                   fancy=True,
                   X=None,
                   label_fontsize=12,
                   ticks_fontsize=8,
                   fontname="Arial")

viz

Random Forest Classification
max depth = 7, criterion = entropy
Number of mislabeled points out of a total 181 points: 56
Accuracy: 0.690608

Read in Future Game Data

Output dataframe head

gameSummaryFutureDB = db['GameSummaryFutureGames']
gamePlayerDataFutureDB = db['GamePlayerDataFutureGames']
playerProjectionsFutureDB = db['PlayerProjectedStatsFutureGames']

future_data = getData(gameSummaryFutureDB, gamePlayerDataFutureDB, playerProjectionsFutureDB)

future_data_df = pd.DataFrame(np.array(future_data), columns=columns)
print("Data count (rows, columns) : " + str(future_data_df.shape))

y_test_future = future_data_df['home win']

future_data_df = future_data_df.drop(['home win'], axis = 1)
future_data_df

Data count (rows, columns) : (11, 34)

	home hr avg	home rbi avg	home tb avg	home avg avg	home ops avg	home so avg	home tpa avg	home obp avg	away hr avg	away rbi avg	...	home pitching pa	away pitching hr	away pitching wpct	away pitching era	away pitching outs	away pitching whip	away pitching h	away pitching ip	away pitching pa	umpire_id
0	11.888889	45.777778	134.111111	0.249222	0.756111	85.111111	352.111111	0.329000	10.333333	37.777778	...	21.0	3.0	1.000	1.08	75.0	0.72	11.0	25.0	90.0	427128.0
1	16.666667	53.777778	160.444444	0.269000	0.771778	73.222222	356.777778	0.320444	8.555556	33.222222	...	463.0	1.0	0.000	10.80	5.0	4.80	5.0	1.2	13.0	521051.0
2	10.666667	31.111111	112.888889	0.240333	0.703000	67.111111	279.666667	0.298444	19.000000	59.333333	...	409.0	7.0	0.500	4.01	101.0	1.10	26.0	33.2	140.0	644760.0
3	12.222222	46.333333	139.222222	0.277778	0.811556	60.111111	333.888889	0.340778	13.777778	42.777778	...	392.0	18.0	0.429	4.10	375.0	1.17	119.0	125.0	517.0	427192.0
4	13.555556	47.444444	148.888889	0.258111	0.784889	80.555556	389.444444	0.345889	15.222222	42.000000	...	577.0	8.0	0.700	2.92	277.0	1.14	78.0	92.1	375.0	484183.0
5	13.111111	42.777778	127.555556	0.230111	0.721667	72.555556	318.777778	0.304667	6.111111	21.333333	...	441.0	1.0	0.000	12.00	9.0	2.67	8.0	3.0	16.0	490319.0
6	10.000000	36.555556	105.777778	0.253222	0.751667	54.222222	256.000000	0.339444	6.888889	26.000000	...	366.0	15.0	0.389	3.39	454.0	1.14	128.0	151.1	622.0	608093.0
7	10.444444	37.000000	124.111111	0.240778	0.693889	60.666667	318.666667	0.299111	6.333333	24.444444	...	628.0	2.0	0.000	1.84	44.0	0.89	11.0	14.2	56.0	427129.0
8	6.111111	17.444444	62.444444	0.215778	0.634444	48.666667	176.888889	0.272222	11.375000	43.250000	...	337.0	6.0	0.750	4.33	156.0	1.35	45.0	52.0	220.0	596740.0
9	13.444444	47.888889	155.222222	0.261000	0.756333	73.000000	388.000000	0.332222	10.555556	36.666667	...	279.0	11.0	0.615	3.18	408.0	1.23	115.0	136.0	562.0	545402.0
10	10.555556	35.222222	109.888889	0.232667	0.717222	70.000000	275.444444	0.300444	5.444444	26.111111	...	506.0	12.0	0.500	4.59	153.0	1.39	47.0	51.0	229.0	503502.0

11 rows × 33 columns

Predict Future Games

future_predictions = clf_rf.predict(future_data_df)

count = 0
for document in gameSummaryFutureDB.find():
    summary_string = document['away_name'] + ' @ ' + document['home_name'] + "  [" + document['game_datetime'] + "]"
    if future_predictions[count] == 1:
        print(summary_string)
        print("     winner: " + document['home_name'])
        #print("     winner: " + document['home_name'] + "  [proba: " + str(round(future_predictions[count],4)) + "]")
    else:
        print(summary_string)
        print("     winner: " + document['away_name'])
        #print("     winner: " + document['away_name'] + "  [proba: " + str(round(future_predictions[count],4)) + "]")
    count = count + 1

# accuracy = 1-(((y_test_future != future_predictions).sum())/len(future_data_df))
# print("Random Forest Classification")
# print("Number of mislabeled points out of a total %d points: %d" % ((len(future_data_df)), (y_test_future != future_predictions).sum()))
# print("Accuracy: %f" % (accuracy))

Chicago White Sox @ Tampa Bay Rays  [2021-08-22T17:10:00Z]
     winner: Chicago White Sox
Detroit Tigers @ Toronto Blue Jays  [2021-08-22T17:07:00Z]
     winner: Toronto Blue Jays
Atlanta Braves @ Baltimore Orioles  [2021-08-22T17:05:00Z]
     winner: Baltimore Orioles
Seattle Mariners @ Houston Astros  [2021-08-22T18:10:00Z]
     winner: Houston Astros
San Francisco Giants @ Oakland Athletics  [2021-08-22T20:07:00Z]
     winner: Oakland Athletics
Washington Nationals @ Milwaukee Brewers  [2021-08-22T18:10:00Z]
     winner: Milwaukee Brewers
Miami Marlins @ Cincinnati Reds  [2021-08-22T17:10:00Z]
     winner: Cincinnati Reds
Pittsburgh Pirates @ St. Louis Cardinals  [2021-08-22T18:15:00Z]
     winner: St. Louis Cardinals
Kansas City Royals @ Chicago Cubs  [2021-08-22T18:20:00Z]
     winner: Kansas City Royals
Philadelphia Phillies @ San Diego Padres  [2021-08-22T19:40:00Z]
     winner: San Diego Padres
Arizona Diamondbacks @ Colorado Rockies  [2021-08-22T19:10:00Z]
     winner: Colorado Rockies